Modern systems for inputting Japanese text use some kind of context-sensitive recognition of phonetic input. (See Ken Lunde's CJKV for more information.) But these require fairly sophisticated linguistic models and lots of CPU power. In the early days, more brute force approaches were tried, such as this giant "keyboard" meant to be used with a stylus to input kanji one at a time. The "keys" are about .5cm square and the table overall is the about same size as tablets still used for CAD and other design tasks.
Input via this tablet really was supported by the Japanese version of the Symbolics system.
Here is some XEmacs code for interfacing to this tablet.
Detail of one section